raptor3x said:3kramd5 said:raptor3x said:That number shouldn't be surprising at all, read noise scales linearly with pixel area.
Shouldn't the A7S have a higher RN than the 1Dx, then?
Code:ISO Measured ISO Read Noise (e-) Pixel size Model EOS-1DX 100 80 38.5 6.9 micron Model A7S 100 80 21.9 8.3 micron
That being said, if someone was offering, I'd take the 1Dx![]()
Sony's using superior technology that gives lower read noise so you can't really compare them directly. A better comparison would be a Nikon D600 with a D800.
Model D800 100 74 4.6 48818 13.4
Model D600 100 79 7.4 76444 13.3
Read noise ratio is (7.4/4.6) ~ 1.6 and pixel area ratio (rough estimate since I'm not accounting for fill factor differences)is (5.9/4.7)^2 ~ 1.57.
There isn't necessarily a scaling of read noise with pixel size. On one level, bigger pixels can accumulate more electrons in their photodiodes, so there is more to be read out, and that could, assuming everything else is identical, result in higher noise. It just takes longer to move that charge around the system, meaning there is more opportunity for noise to be introduced by the myriad of things that can introduce noise.
However, there are also a multitude of factors that play a role in where noise is introduced and how much. Lower readout frequency will usually introduce less noise, regardless. That is usually where improvements in read noise come from. For example, the parallelization of ADC units allows each unit to be reduced in frequency, without losing overall frame readout frequency. Moving from a single ADC to a handful of parallel ADC units allowed Canon, once upon a time, to reduce their read noise a bit. As they have continued to jack up frame rate, those previous gains have been erased in some recent cameras (i.e. the 1D X). Moving to column-parallel ADC allowed Sony to greatly reduce the operating frequency of their ADC units, and paired with some other technological improvements (including fabrication process and materials), they made significant gains in lowering read noise. There are patents for future sensors with per-pixel ADC units, or the true hyper-parallelization of ADC, which allows them to operate at even lower frequency with even further reduced noise.
There are other ways of reducing noise as well. Converting to digital at the earliest opportunity (i.e. CP-ADC or PP-ADC) allows digital transmission, and better error-corrected digital transmission, to be used, thus reducing any further potential for noise to practically nothing (it can still happen, but it's a lot more difficult for a digital signal to become corrupted when you use error-corrected transmission.)
These kinds of improvements can be applied to pixels of any size, so there is not guaranteed to be an increase in read noise with pixel area.
Upvote
0